17 March 2026

SoFi Technologies: Assessment of Muddy Waters Research Short Report

Unbiased analytical assessment. Neither long nor short. For informational and educational purposes only.

Framing Context

Before diving into the substance, two important framing points.

First, Muddy Waters is short SOFI and explicitly disclosed it intends to cover “a substantial majority—possibly all—of its short positions upon publication.” This is not a disinterested academic exercise. Their incentive is maximum stock price disruption on day one. That creates a systematic bias toward worst-case assumptions and inflammatory language.

Second, SoFi’s official response — essentially a blanket denial with no line-by-line rebuttal and a threat of legal action — is conspicuously thin. A company confident in its accounting would rebut the UCC filing evidence directly. It did not. Keefe Bruyette, a constructive sell-side analyst with an Underperform rating and $20 target, said the report “raises some new questions” — a notably measured acknowledgment from a firm that follows SOFI closely.

Accusation 1: Personal Loan Charge-Off Rate Manipulated — Real Rate Is ~6.1%, Not 2.89%

What Muddy Waters Claims

SOFI manipulates its reported charge-off rate through two mechanisms: (a) selling delinquent loans to third-party conduits in the days immediately before they hit the 120-day charge-off threshold, thereby removing them from the charge-off denominator and numerator; and (b) parking defaulted loans off-balance-sheet in unconsolidated VIEs so charge-offs are never recorded on SOFI’s books. MW calculates the adjusted rate at approximately 6.1% versus the 2.89% Q4 2025 reported rate — and argues this inflated input flows directly into SOFI’s Fair Value model to generate $259 million in unwarranted FV gains on Personal Loans, roughly 24.6% of reported EBITDA.

Testing Against Accounting Standards

Under ASC 310-20 and standard bank regulatory guidance (OCC, FFIEC), loans should be charged off when deemed uncollectable and no later than a specified delinquency threshold — for personal loans, typically 120 to 180 days past due. Selling a loan on day 115 to avoid recording the charge-off is not a GAAP violation in isolation; the GAAP loss on the sale gets recorded. However, the issue is what gets fed into the Fair Value model. Under ASC 820 (Fair Value Measurement), SOFI is required to use “market participant assumptions” for default rates. If the observable default rate on its own portfolio is materially higher than what SOFI inputs into its FV model, the model output is unreliable and the fair value carrying amount is overstated.

Evidence Quality

This is MW’s strongest empirical argument. Three pieces of independent corroboration stand out.

First, the mathematical impossibility: the ratio of quarterly charge-offs to the prior quarter’s 30-plus day delinquent balance has repeatedly exceeded 100% starting Q1 2024. That is arithmetically impossible unless loans are being charged off from a population that no longer sits on the reported balance sheet — i.e., from VIEs.

Second, the rating agency data is damning. Fitch has revised its cumulative lifetime default assumptions upward on every single SOFI Personal Loan ABS trust it has reviewed — from 5.00% on SPLT 2023-1 to 8.40% on SCLP 2026-1. Morningstar DBRS sits at 9.26% cumulative on the same SCLP 2026-1 deal. Annualized over the ~1.8 year weighted average life, these translate to approximately 4.7% (Fitch) to 5.1% (DBRS) — meaningfully above SOFI’s 2.80% reported rate, and continuing to rise.

Third, SOFI’s own serviced-portfolio data shows annualized charge-off rates on transferred loans of 6.6% to 13.0% across 2024–2025. Even on SOFI’s own numbers, the serviced-portfolio rate is roughly double the on-balance-sheet rate SOFI uses to calibrate its FV model.

Accusation 2: Student Loan Fair Value — $247M in Unjustified Paper Gains From a Below-Risk-Free Discount Rate

What Muddy Waters Claims

SOFI’s $12.9B student loan portfolio generates approximately $450M per year in “Day 1” Fair Value gains — income recognized at origination before a single payment is collected. The mechanism is a discount rate of 3.89% (FY2025 average) that sits 27–30 basis points below the contemporaneous 10-year US Treasury yield. Using a discount rate below the risk-free rate on consumer credit — which carries default risk, prepayment risk, and essentially zero secondary market liquidity — implies a negative risk premium. This is technically indefensible. MW calculates that discounting at the WAC (5.87%) reduces fair value to exactly par, eliminating the entire $724M cumulative markup.

Testing Against Accounting Standards

ASC 820 requires a reporting entity to maximize the use of relevant observable inputs and minimize unobservable inputs. For a student loan portfolio with limited secondary market trading, the discount rate relies heavily on Level 3 (unobservable) inputs, which gives SOFI substantial discretion. However, the core principle of risk-adjusted discounting is that a risk-bearing instrument cannot rationally be discounted at less than the risk-free rate — that would imply the loans are less risky than US government debt, which is absurd for a consumer lending portfolio running double-digit default assumptions at the ABS level.

Industry practice for student loan FV discounting typically uses a blended rate incorporating the risk-free rate plus a credit spread. SOFI’s auditor (Deloitte & Touche) would need to have approved these discount rate assumptions, which raises questions about their audit procedures.

Evidence Quality

MW’s sensitivity analysis shows clearly: at 4.16% (10-yr Treasury, zero risk premium), FV declines by ~$95M; at 5.39% (Treasury + 123bps conservative credit spread), FV falls below par. The sensitivity math is straightforward and verifiable. The unit economics argument also holds — student loans yielding ~5.9% funded at a blended cost of 6.3% are economically negative before credit and operating costs. Management’s rational motivation to persist in a loss-making business is therefore best explained by Day 1 FV gains feeding into compensation metrics.

Accusation 3: Secured Loans Are Seller-Financed Sales That Fail ASC 860

What Muddy Waters Claims

SOFI’s “Secured Loan” program is structurally a seller-financed whole loan sales program: SOFI lends approximately 80–90% of UPB at ~5% to buyer trusts (Onboard Partners, CSS PL 2023-1, Skyfi Holdings) to fund purchases of SOFI’s own Personal Loans yielding ~13%. SOFI then recognizes these as “sales” under ASC 860, booking Gains on Sale and an inflated Servicing Rights Asset of ~6.2% of UPB (versus the market rate of 0.7–1.7% at comparable lenders). These transactions fail all three ASC 860 sale criteria. If reclassified as secured borrowings, gains on sale would reverse and balance sheet liabilities would increase by $1.7–$1.9B at peak.

Testing Against Accounting Standards

ASC 860-10 establishes three conditions for sale treatment: (1) transferred assets are isolated from the transferor even in bankruptcy; (2) the transferee can pledge or exchange assets; and (3) the transferor does not maintain effective control.

The pledge-back structure documented in the UCC filings fails all three tests definitively. On (1): the buyer trust immediately pledges the purchased loans back to SoFi Bank as collateral for a secured loan — on the same day as the sale. In a bankruptcy, SoFi holds a perfected security interest in the same loans it claims to have sold. Isolation has not occurred. On (2): the buyer cannot freely pledge or exchange the assets because they are already pledged to SoFi as collateral. On (3): SoFi retains the servicing obligation, provides the financing, and can direct the disposition of the assets.

The 6.2% Servicing Rights Asset is also indefensible at market rates. LendingClub books 0.9–1.1%, Rocket ~1.2%, PennyMac ~1.5–2.0%, and SOFI itself used 0.7% on its own LPB sales. A 6.2% SRA — roughly 9x the market rate — generates $87M in fabricated gains on sale that would not exist at a market-rate SRA.

Accusation 4: $312 Million in Unreported Borrowings — Material Misstatement of Financial Statements

What Muddy Waters Claims

Following the SEC comment letter of September 19, 2024, SOFI engineered a rapid reduction in Secured Loan balances. On September 30, 2024 — the last day of Q3 and 13 days after the SEC letter — SOFI transferred CSS PL 2023-1 Pool Loan 2 receivables to its subsidiary SoFi Funding PL VI LLC, which simultaneously pledged them as collateral to JPMorgan Chase Bank, N.A. as senior lender. CFO Lapointe publicly stated on the Q3 2024 earnings call: “we sold $312 million of senior secured loans at a par execution.” The UCC filing directly contradicts this — this was a borrowing, not a sale, with JPMorgan as senior lender and SoFi as mezzanine lender. This $312M borrowing does not appear in Note 9 of SOFI’s financial statements.

Testing Against Accounting Standards

If the transaction is characterized as a borrowing — which the UCC filing structure strongly implies — then ASC 470 (Debt) requires it to be disclosed in the financial statements. The absence of any new line item in Note 9 covering this exposure is what MW characterizes as a material misstatement. All three PL warehouse metrics (collateral, outstanding, capacity) declined Q2 to Q3 2024 — a new $312M facility would have increased all three. “Other Financing Outstanding” was zero in both periods.

The CFO’s statement on the earnings call — that SOFI “sold $312 million of senior secured loans at a par execution” — is directly and specifically contradicted by a public UCC filing dated the same day identifying SoFi Funding PL VI LLC as the borrower and JPMorgan Chase as the senior lender. This is the most serious potential legal exposure in the entire report.

Accusation 5: Loan Platform Business — $251M in Unrecorded Liabilities

What Muddy Waters Claims

SOFI’s LPB, presented to investors as a capital-lite “tollbooth” generating fee income, is actually a wet-funded forward flow arrangement — a form of secured lending where the funder advances money against loans as they are originated. SOFI retains 100% of the first-loss (Class R) residual certificates in SCLP securitizations, services all loans, provides “robust” credit protection as confirmed by counterparty Blue Owl/OWLCX’s own filings, and sponsors all securitizations.

Under ASC 810, a primary beneficiary of a VIE must consolidate it. SOFI’s mischaracterization of its horizontal risk retention as vertical (pro-rata) appears designed to avoid consolidation. MW estimates $251M in unrecorded 2025 LPB liabilities based on the delta between SOFI’s ~6.8% adjusted default rate and OWLCX’s 4.3% contractual default ceiling.

Testing Against Accounting Standards

The horizontal vs. vertical risk retention distinction matters significantly under both ASC 810 and Dodd-Frank risk retention rules (17 CFR 246.4). SOFI holds 100% of the Class R residuals at zero cash consideration — this is definitively a horizontal first-loss position. Claiming it is vertical risk retention on investor calls (“We do a vertical risk retention — 5% of the deal...5% of the A, 5% of the B, 5% of the C”) while Morningstar DBRS presale reports document 100% Class R retention is a factual misrepresentation. The statement on the Q2 2025 earnings call (“We do not retain any credit risk after the transfer”) is directly contradicted by SOFI’s own Class R retention.

Accusation 6: Capitalised Marketing Expenses Permanently Excluded From Adjusted EBITDA

What Muddy Waters Claims

SOFI capitalised approximately $194M in marketing expenses in 2025 (capitalised contract costs balance up 91% YoY to $407.7M). These are capitalised to the balance sheet at origination — bypassing the income statement — then when amortised they are added back as a D&A add-back in Adjusted EBITDA. The result is that marketing expenditure permanently disappears from the metric on which management is compensated. MW deducts 50% ($97M) from EBITDA.

Testing Against Accounting Standards

Capitalising customer acquisition costs is permitted under ASC 340-40 for incremental costs to obtain a contract. However, the “incremental cost” test is specific — only costs that would not have been incurred if the contract had not been obtained qualify. Broad marketing spend (brand advertising, referral bonuses that may not convert) does not meet the test. The 91% YoY growth in capitalised contract costs — far outpacing loan book growth — suggests an aggressive interpretation of what qualifies as capitalisable.

Assessment of Deliberate Management Behaviour — Timing and Obscuring Evidence

This is where the report moves from accounting analysis into intent assessment, and the cumulative weight of evidence is troubling. Six specific behavioural patterns stand out.

The totality of these six patterns — timed balance sheet actions, quiet disclosure changes not flagged to investors, parallel verbal characterisations that contradict UCC filings, program lifecycle transitions that track SEC scrutiny, and compensation-maximising accounting choices at every node — is more consistent with deliberate financial engineering than with a series of coincidental decisions.

SoFi’s Response — An Assessment

SoFi’s official statement is conspicuously non-specific. It calls MW’s report “inaccurate and misleading” and “designed to deceive investors,” emphasises regulatory oversight (Federal Reserve, OCC), and threatens legal action. It offers zero line-by-line rebuttal of the UCC filing evidence, the Note 9 analysis, the rating agency default assumption data, or the CFO’s earnings call statement versus the September 30, 2024 UCC filing.

For a company “with strong confidence in the integrity of its financial reporting,” the absence of a direct rebuttal to specific, verifiable documentary evidence is telling. Threatening legal action against a short seller while declining to address the substance is a classic defensive posture that typically signals more concern about further disclosure than about winning a defamation case.

Overall Verdict

What Muddy Waters Gets Right, With High Confidence

The charge-off rate understatement (Accusation 1), the sub-risk-free student loan discount rate (Accusation 2), and the ASC 860 failure on Secured Loan sales (Accusation 3) are each well-supported by documentary evidence and technically sound accounting analysis. These three together would, if validated by a regulator or auditor, require material restatements of revenue, EBITDA, and balance sheet. The $312M unrecorded borrowing (Accusation 4) is the most legally serious — a specific, verifiable, and apparently unrebutted contradiction between a CFO’s earnings call statement and a public UCC filing.

Where Muddy Waters Likely Overstates

The 90% EBITDA haircut (from $1,054M to $103M) assumes all adjustments are simultaneously correct at their most aggressive estimates, with no offsets. In practice, some of the LPB liability estimate ($251M, Accusation 5) and the marketing capex haircut ($97M, Accusation 6) involve more contested assumptions. A reasonable central estimate of EBITDA overstatement is probably in the $400–600M range rather than $950M — still material, still concerning, but not the apocalyptic scenario MW implies.

What This Means for Investors

The report raises questions that SOFI must answer specifically: Does the $312M JPMorgan facility appear anywhere in the financial statements? Has Deloitte reviewed the Secured Loan seller-financing structure in full? What is SOFI’s basis for a discount rate below the Treasury yield on consumer credit?

The investment thesis on SOFI has shifted from a valuation debate to an accounting integrity debate, which is a qualitatively different risk category. Until SOFI provides substantive, document-specific rebuttals — particularly on the $312M and ASC 860 issues — the risk profile is materially elevated. CEO Noto buying $500K of stock on the day of publication is a positive signal but does not constitute rebuttal evidence. The appropriate institutional response to this report is to demand clarification on Accusations 3 and 4 before forming a position.